INDEX

Explanations

drew lacked earned to send why noboby

The neuron strongly activates on verbs in their continuous/participial (“-ing”) form, flagging ongoing actions in the text.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 Meksiku

-1.70

ↀ

-1.48

にほんブログ村

-1.43

ſi

-1.41

鸲

-1.39

ではでは

-1.38

pondre

-1.38

 우리

-1.38

ではまた

-1.38

我跟

-1.37

POSITIVE LOGITS

 only

1.30

 should

1.28

 could

1.21

1.20

景观

1.16

 upon

1.16

1.14

to

1.13

 would

1.12

one

1.08

Activations Density 0.185%